On Sun, Jul 03, 2011 at 11:27 -0700, Linus Torvalds wrote:
> That patch is entirely insane. No way in hell will that ever get merged.

Sure, this is just an RFC :)  I didn't think about proposing it as a
patch as is, I tried to just show how/what checks it introduces.


> copy_to/from_user() is some of the most performance-critical code, and
> runs a *lot*, often for fairly small structures (ie 'fstat()' etc).
> 
> Adding random ad-hoc tests to it is entirely inappropriate. Doing so
> unconditionally is insane.

That's why I've asked whether it makes sense to guard it with
CONFIG_XXX, defaults to =n.  Some distributions might think it makes
sense to enable it sacrificing some speed.

Will do.


> If you seriously clean it up (that at a minimum includes things like
> making it configurable using some pretty helper function that just
> compiles away for all the normal cases,

Hm, it is not as simple as it looks at the first glance - even if the
object size is known at the compile time (__compiletime_object_size), it
might be a field of a structure, which crosses the slab object
boundaries because of an overflow.

However, if interpret constants fed to copy_*_user() as equivalent to
{get,put}_user() (== worry about size argument overflow only), then it
might be useful here.


>    if (!slab_access_ok(to, n) || !stack_access_ok(to, n))

OK :)


Thanks!

-- 
Vasiliy Kulikov
http://www.openwall.com - bringing security into open computing environments